• fatigue curves;
  • iterative quantile regression;
  • genetic algorithms;
  • structural risk minimization;
  • censored data;
  • general approximate cross-validation error

Accurate prediction of fatigue failure times of materials such as fracture and plastic deformation at various stress ranges has a strong bearing on practical fatigue design of materials. In this study, we propose a novel genetic-based iterative quantile regression (GA-IQR) algorithm for analyzing fatigue curves that represent a nonlinear relationship between a given stress amplitude and fatigue life. We reduce the problem to a linear framework and develop the iterative algorithm for determining the model coefficients including unknown fatigue limits. The procedure keeps updating the estimates in a direction to reduce its resulting error. Also, our approach benefits from the population-based stochastic search of the genetic algorithms so that the algorithm becomes less sensitive to its initialization. Compared with conventional approaches, the proposed GA-IQR requires fewer assumptions to develop fatigue model, capable of exploring the data structure in a relatively flexible manner. All procedures and calculations are quite straightforward, such that the proposed quantile regression model has a high potential value in a wide range of applications for exploring nonlinear relationships with lifetime data. Computational results for real data sets found in the literature present good evidences to support the argument. Copyright © 2012 John Wiley & Sons, Ltd.